Bayesian Hausman Test
Bayesian Hausman Specification Test · Also known as: Bayesian specification test, Bayesian endogeneity test, Bayesian FE vs RE test, Bayesian Durbin-Wu-Hausman
The Bayesian Hausman test is a Bayesian reformulation of Hausman's (1978) classical specification test, used to assess endogeneity or to choose between fixed effects and random effects panel models. Instead of a chi-squared test statistic, it uses posterior model probabilities or Bayes factors to compare competing specifications, fully incorporating prior uncertainty about model parameters.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Bayesian Hausman test when you face the fixed-vs-random effects choice in panel data analysis and want to incorporate prior information or report full probabilistic evidence rather than a binary p-value decision. It is particularly useful when the classical Hausman test is inconclusive (borderline p-values), when the sample is small and asymptotic chi-squared approximations are unreliable, or when you are already working within a Bayesian estimation framework. Do not use it as a drop-in substitute for the classical test without understanding the sensitivity of Bayes factors to prior specification — poorly chosen priors can dominate the inference, especially in small samples.
Strengths & limitations
- Provides a continuous measure of evidence (posterior odds or Bayes factor) rather than a binary accept/reject decision.
- Naturally incorporates prior information about model plausibility, which can be valuable when domain knowledge is available.
- Does not rely on asymptotic approximations, making it more reliable than the classical test in small samples.
- Produces posterior model probabilities that directly quantify uncertainty about the correct specification.
- Coherently extends to multi-model comparison beyond the binary FE vs RE choice.
- Bayes factors are sensitive to the choice of priors on parameters, particularly diffuse or improper priors, which can make results unreliable.
- Marginal likelihood computation is often analytically intractable and requires numerical integration (e.g., MCMC, bridge sampling), increasing computational cost.
- Interpretation of Bayes factors requires familiarity with the Bayesian framework; results are less familiar to audiences trained in classical econometrics.
- No universally accepted software routine exists for the Bayesian Hausman test, unlike the classical version available in every econometrics package.
Frequently asked
How does the Bayesian Hausman test differ from the classical Hausman test?
The classical Hausman test computes a chi-squared statistic comparing two estimators and yields a p-value. The Bayesian version computes the marginal likelihood of the data under each model and produces a Bayes factor or posterior model probabilities — a continuous, probabilistic measure of evidence rather than a binary significance decision.
What is a Bayes factor and how do I interpret it?
The Bayes factor is the ratio of marginal likelihoods of two models: BF = p(y|M1) / p(y|M2). Values above 1 favor M1; above 3 is considered moderate evidence; above 10 is strong. Values below 1 favor M2. It updates prior odds into posterior odds.
Do I need MCMC to run the Bayesian Hausman test?
Often yes — marginal likelihoods require integrating over parameter spaces, which is rarely tractable analytically. MCMC with bridge sampling or harmonic mean estimators is typical. In simple conjugate models, analytical marginal likelihoods may be available.
Can I use this test to detect endogeneity more broadly, not just for FE vs RE?
Yes. The Bayesian Hausman principle generalizes to any pair of models where one estimator is consistent under a broader model and another is efficient only under a restricted model. It applies to instrumental variables, simultaneous equations, and other endogeneity settings.
How sensitive are the results to my prior choices?
Bayes factors can be quite sensitive to priors, especially with small samples. Always conduct prior sensitivity analysis by varying the prior hyperparameters and reporting how the Bayes factor changes — this is standard practice in Bayesian model comparison.
Sources
- Hausman, J. A. (1978). Specification tests in econometrics. Econometrica, 46(6), 1251–1271. DOI: 10.2307/1913827 ↗
- Lancaster, T. (2004). An Introduction to Modern Bayesian Econometrics. Blackwell Publishing. ISBN: 978-1405117203
How to cite this page
ScholarGate. (2026, June 3). Bayesian Hausman Specification Test. ScholarGate. https://scholargate.app/en/econometrics/bayesian-hausman-test
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bayesian Fixed Effects ModelEconometrics↔ compare
- Bayesian Panel Data AnalysisEconometrics↔ compare
- Bayesian Random Effects ModelEconometrics↔ compare
- Fixed Effects ModelEconometrics↔ compare
- Panel Hausman TestEconometrics↔ compare